Am Freitag, 12. Juni 2009 17:05:33 schrieb Jeremy Olexa:
> > It seems that whole process could be sped up by hosting binary
> > packages on one central server (Binary host). Obviously various versions
> > of the same package would be created and therefore unique names could be
> > created by using some metadata to create hash part of filename. On a
> > first thought I would use USE flags and DEPEND as metadata to hash.
>
> This is a cool aspect of the project, I hope you can work with solar
> and zmedico to improve binpkgs. USE flags seem to be the trouble spot
> of binpkgs.

Maybe this Bug could help with that: 

"Include more info about a binpkg"
- http://bugs.gentoo.org/show_bug.cgi?id=150031#c7

I wanted to tackle that myself some time ago but got stuck at trying to find 
my way within portage so I postponed it a few times... 

It would only need three things: 


* Find SLOT and USE for any ebuild (SLOT is everything which affects all 
packages on one machine, USE are the *active* USE-flags of the package). 

* Save binpackages in a directory structure which contains a hash over the 
SLOT and a hash over the active USE flags. 

* Locate binpackages via the same hashes. 


The path to a binpackage would then look like this: 

$PKGDIR/$CATEGORY/$PN/$SLOTS_HASH/$USE_HASH/python-2.5.2-r8.tbz2


You can make the USE hash more user-friendly by only hashing a USE string, if 
the unhashed string would be longer than the hashed string. 

As hash you can simply use sha.sha() - maybe in BASE32 encoding to make it 
URL- and filename-safe. 

--- !python
# hashing a USE string
from sha import sha
from base64 import b32encode

hash = sha()
use_string = "STRING OF THE ACTIVE USE FLAGS"
# if the string is longer than a base32 encoded sha1 hash, 
# hash it to have an upper bound on its length
if len(use_string) > 32: 
	hash.update(use_string)
	use_string = b32encode(hash.digest())
...

One missing piece is a nice human readable delimiter for the USE flags in the 
USE string which can safely be used in URLs. If there is none, always hashing 
USE flags is a simple fallback option. 

Since "-" and "_" are already being used in USE-flags I don't know about more 
possible delimiters - except just using the space and escaping it in URLs as 
%20.  

The bash won't like that (require escaping), but since portage is written in 
Python and ebuilds don't need to access binpackages directly that should be a 
minor problem. 

Best wishes, 
Arne

--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- 
   - singing a part of the history of free software -
              http://infinite-hands.draketo.de